Visualizing Uncertainty

March 3-4, 2005

NAS Workshop: Visualizing Uncertainty

Next steps

lHeighten awareness of the problem among
public, professionals, researchers & developers

lSupport multi-valued data representation standards

lExplore techniques for each data type

lDevelop guidelines for implementers:

–Data formats

–Interactive interfaces

–Visual presentations

lDevelop human performance evaluation methods

lPublish benchmark datasets & evaluation metrics

lForm guidelines for how to
propagate/integrate uncertainty markers


	1.Multivalued data sets are N times larger than uni-valued data. Visualization may not be the
	right/immediate solution. We need to do feature extraction first – analyze then visualize.
	Feature extraction will also need to work with multivalued data.
	2. We just scratched the surface in terms of visualizing multivalued data.
	Need to explore other ways of visualizing it;
	3. Need to train/evaluate users in these visualization.
	Choice of operators and metrics is large and can be drawn from different
	fields e.g. signal processing, info theory, statistics, etc.
	Need to know which is the appropriate one to use for different context.
	4.The current multivalue representation does not preserve the spatial
	correlation information in the different realizations. The visualization
	and analyses algorithms can try to recover that info. This requires
	working with a matrix of values rather than a column of values for
	each scalar variable at each location in space!